disk mirror and power/hardware failures - is it safe enought ?

[unknown]

Hi everybody,
I've setup mirrored disks
(on separate physycal devices - for the case of one of
devices will go broken)
to be safe on power failure. I used syncronous mirrors
( when writes are sequential, not parallel) -
so that on power down, reset etc. at least one of two copies
(old or new one) as I expect will be completely written and
thus
consistent and ASE will be able to recover data 100%.

Q: Is it a good idea, and is it really safe approach ?

Also, once 250days uptime server got SCSI controller hard
failed (I guess so
because kernel reported hard errors on many devices, but
after reboot all is ok).
And one of DBs had fatal error 695 ( actually only one
table, fortunately I was able to
bcp dump all others). The writeup says:
"The error may be due to an Adaptive Server problem but
could also be caused
by problems such as overlapping partitions or hardware
errors"

Q: Have any ideas how to be safe even on this ? ( exept
having backups :) - I've
a huge databases, total 4Tb of disks - so backing up of
everything is a problem )

Now I'm planning move to new hardware, where I can setup
different hardware RAID
configurations + software (ASE) mirroring; I want to use the
above software mirror
setup + RAID0 (striping) to get maximum HDD throughput

Q: What is your opinion (experience) on such setups ?

Thank in advance for any advices,
Serge

[A. M.]

Post by unknown
I've setup mirrored disks
(on separate physycal devices - for the case of one of
devices will go broken)
to be safe on power failure. I used syncronous mirrors
( when writes are sequential, not parallel) -
so that on power down, reset etc. at least one of two copies
(old or new one) as I expect will be completely written and
thus consistent and ASE will be able to recover data 100%.
Q: Is it a good idea, and is it really safe approach ?

From your statements, I suspect you are referring to
ASE's own disk mirroring. This is really an option
of last resort. Try to use mirroring provided by
hardware or your OS instead.

Post by unknown
Also, once 250days uptime server got SCSI controller hard
failed (I guess so
because kernel reported hard errors on many devices, but
after reboot all is ok).
And one of DBs had fatal error 695 ( actually only one
table, fortunately I was able to
"The error may be due to an Adaptive Server problem but
could also be caused
by problems such as overlapping partitions or hardware
errors"
Q: Have any ideas how to be safe even on this ? ( exept
having backups :) - I've
a huge databases, total 4Tb of disks - so backing up of
everything is a problem )

If you really have 4 TB of disk (and presumably
data), you really should be using either OS-level
or hardware-level RAID. Raid 1+0 is ideal. Striping
for performance and mirroring for security.

Backups shouldn't pose a problem is some simple
design rules are followed. If you have unrelated
tables, placing them in separate databases helps
reduce the backup issue (assuming large tables
are distributed fairly evenly across these databases).

Post by unknown
Now I'm planning move to new hardware, where I can setup
different hardware RAID configurations + software (ASE)
mirroring;

You don't need ASE's primitive mirroring if you
have RAID.

Post by unknown
I want to use the above software mirror
setup + RAID0 (striping) to get maximum HDD throughput
Q: What is your opinion (experience) on such setups ?

Use RAID 1+0. For critical devices you could do
3 way mirroring (or more) but I've never found
the need to do more than 2 way mirroring (disks
are fairly reliable). Small, lightly used ASE
devices (like master, sybsystemprocs) don't
benefit from striping so just mirror those
devices.

-am © 2007